Samsung to Manufacture Tesla’s Custom AI Chips by Late 2027 in $16.5 Billion Foundry Deal
The multi-year contract marks Tesla's deepening vertical integration in semiconductor supply chains and Samsung's breakthrough in automotive-grade foundry capacity.
Samsung Electronics confirmed today that volume production of Tesla’s custom AI chips will begin in the second half of 2027 at its Texas facility, formalizing a $16.5 billion contract that runs through December 2033. The deal, announced by Han Jin-man, Samsung’s President and Head of Foundry Business, at a shareholders meeting, represents Samsung’s most significant automotive foundry win and Tesla’s latest move to reduce dependence on third-party chip suppliers.
The agreement positions Samsung’s Taylor, Texas fabrication facility—expected to begin operations in the second half of 2026—as Tesla’s dedicated production line for next-generation autonomous driving processors. According to Reuters, Samsung expects the Tesla contract to generate approximately 2.5 trillion won ($1.8 billion) annually, representing roughly 10% of Samsung’s projected foundry revenue through the end of the decade.
Breaking TSMC’s Automotive Stranglehold
The deal fractures TSMC’s dominance in Automotive semiconductor manufacturing, where the Taiwanese giant holds 64-66% of global foundry market share. Samsung currently commands 7-13% share, per TrendForce, with Intel trailing below 5%. Tesla’s decision to split production between Samsung (for custom AI Chips) and TSMC (which manufactures components for Tesla’s AI5 processor) reflects a deliberate strategy to avoid single-supplier risk in a geopolitically volatile semiconductor landscape.
“The Tesla contract suggests that Samsung’s yield on its advanced nodes, particularly 2-nanometer technology, is finally reaching viable levels. It also signals renewed customer trust and a possible shift in market momentum.”
— Semiconductor executive close to the matter, speaking to KED Global
Samsung’s 2nm foundry capacity is projected to more than double by the end of 2026, with yield rates hitting 55-60% as of November 2025—a critical threshold for automotive-grade manufacturing, which demands higher reliability than consumer electronics. The Tesla anchor tenant model mirrors TSMC’s relationship with Apple, providing Samsung with predictable revenue to justify capital expenditure on advanced nodes.
Tesla’s Semiconductor Roadmap Accelerates
Tesla’s custom chip strategy now spans three generations. The AI4 processor, manufactured on Samsung’s 7nm process, currently powers the company’s Full Self-Driving (FSD) system. The AI5 chip, delayed to late 2026 or early 2027 production, uses TSMC’s 4N process (a 5nm-class node). The AI6, scheduled for late 2027 volume production at Samsung’s Texas facility, will leverage more advanced process technology—likely Samsung’s 2nm or 3nm nodes.
Tesla CEO Elon Musk has stated the company is “not about to replace Nvidia” but uses both custom silicon and Nvidia GPUs in combination, according to CNBC. While Nvidia’s Drive Thor chip targets the broader automotive AI market using TSMC’s 4N process, Tesla’s vertical integration strategy reduces per-unit costs and allows tighter integration between hardware and FSD software.
On March 14, 2026—four days before Samsung’s confirmation—Tesla announced Terafab, an in-house semiconductor fabrication project targeting $20-25 billion in investment. The announcement signals Tesla’s long-term ambition to bring chip manufacturing fully under its control, though the Samsung deal ensures intermediate-term supply through 2033 as Terafab development progresses.
Implications for Nvidia and Automotive Supply Chains
The Samsung-Tesla partnership intensifies pressure on Nvidia’s automotive AI revenue stream. While Nvidia dominates training infrastructure for autonomous systems, Tesla’s custom inference chips reduce dependency on Nvidia’s Drive platform for in-vehicle computation. Electrek analysis shows Tesla AI4 already offers competitive compute density against Nvidia Thor, with AI6 expected to widen that gap.
- Samsung secures stable foundry revenue (~$1.8B annually) as it competes with TSMC for advanced node leadership
- Tesla reduces per-chip costs and supply chain exposure through vertical integration
- Nvidia faces margin pressure in automotive AI inference as major OEMs pursue custom silicon
- Geopolitical diversification accelerates as US-based fabs (Samsung Texas, Intel, potential Tesla Terafab) challenge Taiwan-centric Supply Chains
The broader semiconductor foundry market is projected to grow from $202 billion in 2026 to $263.1 billion by 2034, representing a 3.4% compound annual growth rate, according to Fortune Business Insights. Automotive chips, driven by EV adoption and autonomous driving requirements, represent one of the fastest-growing segments within that market.
What to Watch
Samsung’s ability to meet Tesla’s late 2027 volume production timeline depends on continued yield improvements at its 2nm and 3nm nodes. Any delays would force Tesla to either extend reliance on AI5 processors or shift additional capacity to TSMC, undermining the supply diversification strategy. Monitor Samsung’s quarterly foundry revenue disclosures for early signals of production ramp issues.
Tesla’s Terafab announcement suggests the Samsung contract may represent a bridge rather than a permanent arrangement. If Terafab achieves target timelines (likely late 2020s for volume production), Samsung could lose its anchor tenant by the mid-2030s. The Taylor facility’s long-term viability hinges on winning additional automotive or AI customers beyond Tesla.
Nvidia’s response to Tesla’s vertical integration will shape competitive dynamics in automotive AI. Watch for Drive Thor pricing adjustments, expanded partnerships with traditional OEMs (GM, Ford, BMW), or acceleration of Nvidia’s own custom foundry agreements to offset revenue pressure from Tesla’s in-house chip strategy.